Our brains store a tremendous amount of knowledge about the world. We use this knowledge constantly in our daily lives. We use it to make sense of objects, words, actions, numbers, people and music. This stored knowledge is referred to as semantic memory. While “semantics” is a term that refers to meaning specifically in language, “semantic memory” addresses the broader notion of our long-term accumulated knowledge of the world. One of the big questions that we investigate at the Penn FTD Center is: How is semantic memory represented in the brain? These studies cover a wide range of domains and processes within the semantic memory system, and illustrate the large functional network that underlies the representation of knowledge in the brain. By examining these processes in both healthy adults and patients with cognitive impairments, we are able to better understand the neural basis of semantic memory and how it is affected by neurodegenerative disease.
Many theories about the neural basis of semantic memory have focused on the representation of object and action concepts. Examples of object concepts include things like “apple” and “hammer.” Examples of action concepts include “kick” and “grasp.” It is proposed that object and action concepts are represented in or near regions of the brain that support perception and action. Thus, the semantic representation of “apple” includes visual brain regions involved in color and shape perception. Our concept of “apple” also may involve gustatory regions where the taste of an apple may be represented and auditory brain regions where the crunch of an apple may be represented. Similarly, the semantic representation of “kick” includes motor brain regions that control leg movements and visual-motion regions for perceiving the action associated with “kick.” In other words, the semantic memory system appears to overlap in part with the sensory and motor systems of the brain. Hence, these theories have been referred to as sensory-motor theories of semantic memory. This perspective has strong intuitive appeal because it links our direct experience of the world through sensation and action with our stored knowledge of the world in semantic memory.
In the Penn FTD Center, we investigate this hypothesis from a number of perspectives. As in our other areas of research, we typically seek converging evidence from both fMRI experiments in healthy adults and structural MRI experiments in patients with cognitive impairments resulting from neurodegenerative disease. Using this approach, we have demonstrated a critical role for visual, auditory, and motor regions of the brain in the representation of object and action concepts.
There are many other brain regions besides the sensory and motor cortices. One of the shortcomings of sensory-motor theories of semantic memory is that they often pay little attention to “heteromodal” regions of the brain. Heteromodal brain regions are those where numerous modalities of sensory and motor information converge. Not surprisingly, such heteromodal regions are frequently implicated in neuroimaging studies of semantic memory. These regions include inferior parietal, lateral temporal, and lateral prefrontal cortices. All of these regions have reciprocal anatomic connections with nearly all sensory and motor regions of the brain. This anatomic arrangement makes these regions well suited to performing high-level processes in semantic memory, such as binding together the many features associated with a concept or performing goal-directed semantic memory tasks.
Researchers at the Penn FTD Center have developed several lines of investigation that characterize the role of heteromodal brain regions in semantic memory. This work points to regions of the inferior parietal and lateral temporal cortices in representing all categories of concepts, regardless of their specific sensory and motor associations. These parietal and temporal regions also appear to have a role in combining information from across multiple concepts, a process we do every time we read a sentence and combine the words together in a meaningful way. Additionally, our work has characterized the role of lateral prefrontal regions in goal-directed processes in semantic memory. For example, lateral prefrontal regions appear to be involved in the retrieval of specific information from semantic memory, and in the use of logic to make inferences from semantic information. Finally, we have attempted to dissociate known concepts represented in the brain from the neural architecture needed for the acquisition and representation of novel concepts.
At the Penn FTD Center, we also investigate several other semantic domains besides object and action knowledge. For example, we are interested in understanding the neural basis of abstract concepts, such as “truth.” Abstract concepts do not have strong sensory or motor associations, and they present a challenge to sensory-motor theories of semantic memory. Our work suggests that abstract concepts rely heavily on representations in heteromodal regions.
We are also actively investigating the neural representation of number and quantity concepts. These concepts involve collections or amounts of substances regardless of the specific identity of the substances. This work may implicate brain regions that are important for magnitude. Our research has shown that a very common class of words – quantifiers – appears to depend on number knowledge. A quantifier is a word like “many,” “most” or “less.” Even though quantifiers are words, functional imaging studies of quantifiers demonstrate activations in non-word areas that are important for appreciating number knowledge. Likewise, non-aphasic patients who do not have a language disorder but have difficulty with numbers also have difficulty understanding quantifiers.
We have been studying the neural basis for appreciating the meaning of music as well. In patients with profound difficulty understanding objects and object names, we find striking preservation of the ability to interpret the meaning of music.